Table 5 The experimental results of treatment plans as the first part of Phase I predictions. CC denoted Clomiphene Citrate, OCP denoted Oral Contraceptives, GnRHant denoted GnRH Antagonists, Trigger denoted Trigger Medication. We uses the three following metrics for Phase I predictions (split by the symbol | from left to right): average precision score (AP), area under the receiver operating characteristic curve (AUROC), and top 2 accuracy score (top-2). The category Folitropin had an extremely imbalance label set so that all of the validation data being predicted as the dominate class by AdaBoost yet achieved the best AP (93.0%).

Edwards-Pro

74.7\({\textbf {|}}\)90.6\({\textbf {|}}\)79.0

88.3|93.6|88.1

88.4|91.1|92.7

87.3|95.9|89.3

80.7|82.0|81.6

84.6|91.3|87.3

79.6|82.2|79.8

Edwards

70.2|88.5|75.2

74.5|87.8|76.5

85.5|87.1|92.3

86.9|95.5|84.9

71.4|80.9|73.3

83.7|91.1|85.0

75.6|81.5|76.2

Seq2Seq

65.1|82.8|70.3

86.3|91.8|87.5

80.7|82.2|82.5

83.1|84.6|83.0

73.4|79.0|74.6

79.8|88.6|84.6

73.0|77.6|72.4

Nearest Neighbors

31.7|36.0|32.5

85.6|61.3|87.8

57.3|56.1|70.7

61.1|54.6|71.2

44.2|51.7|47.8

49.6|52.6|54.9

40.6|51.6|45.1

SVM

33.9|47.8|38.8

87.1|68.3|87.8

66.8|68.0|66.2

76.6|70.3|77.0

60.3|58.6|62.0

67.0|69.1|70.7

61.5|63.8|65.4

AdaBoost

29.9|41.7|37.4

93.0|67.7|94.1

68.9|63.4|84.8

75.9|62.5|87.7

61.3|55.7|67.8

65.7|61.3|68.2

64.6|56.9|71.9

Random Forest

27.9|36.6|37.9

91.0|74.8|92.3

68.5|73.0|83.4

76.1|68.8|86.8

61.8|63.0|69.4

65.5|64.8|65.9

61.8|63.0|69.4

Naive Bayes

37.5|45.6|39.8

70.8|58.9|73.9

58.3|60.1|67.0

51.7|60.2|63.1

30.1|52.9|20.0

43.4|57.9|35.1

40.7|54.3|48.7

Decision Tree

36.5|45.0|37.8

87.3|78.7|93.4

67.3|71.2|80.8

76.0|71.0|85.0

62.2|61.4|69.4

65.8|64.0|67.4

64.7|59.4|72.8

Neural Net

35.4|40.0|32.1

86.5|74.8|87.6

63.0|66.6|73.8

71.5|62.9|79.8

61.2|60.0|66.8

63.7|62.7|64.5

61.2|60.1|66.8